A public research preview
MetaAware
Human-directed, evidence-bound governance for AI-assisted software delivery.
MetaAware helps developers understand what their project state supports, what remains uncertain, and what evidence is needed before the next consequential action.
ObserveExplainDecideHuman
authority
authority
Project phaseIncubation
Source repositoryPrivate
CLINot yet implemented
Current focusRead-only prototype design
Last reviewed2026-09-13
The governance gap
Implementation capacity is growing faster than governance capacity.
AI lowers the cost of producing software change. It does not automatically give a developer equivalent experience in architecture, testing, security, integration, or release decisions.
MetaAware explores how stronger governance can remain affordable and useful before that expertise has been fully acquired.
Responsibility model
AI can assist governance. It cannot grant itself authority.
Humans constitute→Machines observe→Rules decide→AI explains→Humans remain accountable
Humans constitute. Machines observe. Rules decide. AI explains. Humans remain accountable and retain final authority.
The recursive chain must terminate in inspectable evidence and real outcomes, not in one AI system agreeing with another.
Four principles
Governance must work before expertise has been acquired.
Every automated judgment should increase human understanding, not merely compliance.
Governance authority remains human, and judgments must be evidence-bound.
Conceptual output contract
A finding should make the path forward visible.
Not yet implemented in the CLI.
BLOCKED
REVIEW_STALE- Observed fact
- Current HEAD differs from the revision that was reviewed.
- Why this matters
- The existing review does not cover the latest change.
- Next safe step
- Review the difference and bind the new review to current HEAD.
- Verify recovery
- Confirm the new review record identifies the current revision.
- Epistemic status
- Observed Git fact + project declaration
This finding does not prove that the new code is correct, secure, or ready for users.
Evidence so far
12 of 12 expected conditions detected by a deterministic raw-Git oracle.
Twelve governance scenarios derived from two development contexts were rebuilt as disposable synthetic Git repositories. The tested default command surfaces of gitgov 3.0.0 and govctl 0.14.1 did not directly identify those scenario-specific conditions.
Inspect the method and limitations →Medium- and long-term horizon
From delivery evidence to security governance.
MetaAware is designed to extend beyond Git workflow without becoming another vulnerability scanner. The accepted direction is to govern how security evidence is connected to source, deployment state, policy, uncertainty, approval, and recovery.
Secure delivery
The first post-MVP experiment will connect scanner, dependency, SBOM, CI, and infrastructure evidence through a versioned policy pack.
Runtime and network posture
Read-only adapters may later observe cloud, network, and runtime state. Missing external evidence remains unknown.
Identity and business integrity
Account and payment governance will require project-declared invariants, executable scenarios, and independent domain review.
Verified progress
Build public understanding before claiming public proof.
MetaAware has an accepted governance philosophy, product boundary, technical direction, external-validation program, and public/private evidence boundary.
Accepted
Public research website design
This public preview implements the accepted communication design. It does not implement the governance CLI.
Open the progress ledger →Founder perspective
Shaped by AI safety evaluation and hands-on AI-assisted development.
MetaAware was founded at the intersection of data-driven risk controls, AI safety evaluation, and direct experience coordinating software delivery with coding agents.
MetaAware is an independent project. References to prior professional experience describe the founder's background and do not imply employer sponsorship or endorsement.
About the founder perspective →